State Transition Graph-Based Spatial-Temporal Attention Network for Cell-Level Mobile Traffic Prediction.

Area of Science:

• Computer Science
• Telecommunications Engineering
• Data Science

Background:

• Efficient mobile traffic prediction is crucial for optimizing network resources and enhancing user experience.
• Existing methods often struggle to fully capture the complex spatial-temporal dynamics inherent in mobile traffic data.

Purpose of the Study:

• To propose a novel spatial-temporal attention network (STG-STAN) for cell-level mobile traffic prediction.
• To effectively exploit underlying spatial-temporal dynamic information from historical mobile traffic data using state transition graphs.

Main Methods:

• Constructing state transition graphs to identify semantic context and patterns in historical data.
• Employing graph convolutional networks (GCNs) for spatial information aggregation within state transition graphs.
• Utilizing a temporal extraction module to capture the dynamic evolution of state transition graphs over time.
• Integrating the spatial-temporal attention network with a long short-term memory (LSTM) module for enhanced prediction accuracy.

Main Results:

• The STG-STAN effectively captures spatial-temporal information embedded within state transition graphs.
• Experimental results demonstrate superior performance of STG-STAN compared to several baseline methods.
• The model shows significant improvements in mobile traffic prediction accuracy.

Conclusions:

• The proposed STG-STAN offers a robust approach for cell-level mobile traffic prediction.
• Leveraging state transition graphs enhances the model's ability to understand complex data dynamics.
• The STG-STAN provides a promising direction for future research in mobile network optimization.